MCP server for AI agent security -- input validation, prompt injection detection, PII redaction, policy enforcement, rate limiting, and audit logging